Large language models can now use external tools - checking databases, calling APIs, processing payments. This capability, called function calling, transforms a chatbot into an agent that completes multi-step tasks autonomously.

This week, I dove into RAG systems. Some optimize for speed, others for precision, and others for multi-source synthesis. Here's a framework for choosing the right architecture.

Here's what happened when I decided to ship a feature myself.

I built a hallucination detector with four dimensions to test my AI tutor. The results revealed surprising insights about enrichment vs. true hallucination.

I built keyword search, semantic search, and hybrid search for my language learning app. Here's what actually worked.

From Mary Oliver to modern AI: understanding the mechanism that powers every breakthrough in language AI

A photographic journey through Cerritos, California, uncovering remnants of drive-through dairy farms that once thrived in the 1950s